Arch Barbara N, Blair Joanne, McKay Andrew, Gregory John W, Newland Paul, Gamble Carrol
Department of Biostatistics, The University of Liverpool, Liverpool, L69 3BX, UK.
Alder Hey Children's NHS FT, East Prescott Road, Liverpool, L12 2AP, UK.
Trials. 2016 Oct 24;17(1):517. doi: 10.1186/s13063-016-1640-6.
Glycated haemoglobin (HbA1c) is an important outcome measure in diabetes clinical trials. For multicentre designs, HbA1c can be measured locally at participating centres or by sending blood samples to a central laboratory. This study analyses the agreement between local and central measurements, using 1-year follow-up data collected in a multicentre randomised controlled trial (RCT) of newly diagnosed children with type I diabetes.
HbA1c measurements were routinely analysed both locally and centrally at baseline and then at 3, 6, 9 and 12 months and the data reported in mmol/mol. Agreement was assessed by calculating the bias and 95 % limits of agreement, using the Bland-Altman analysis method. A predetermined benchmark for clinically acceptable margin of error between measurements was subjectively set as ±10 % for HbA1c. The percentage of pairs of measurements that were classified as clinically acceptable was calculated. Descriptive statistics were used to examine the agreement within centres. Treatment group was not considered.
Five hundred and ninety pairs of measurement, representing 255 children and 15 trial centres across four follow-up time points, were compared. There was no significant bias: local measurements were an average of 0.16 mmol/mol (SD = 4.5, 95 % CI -0.2 to 0.5) higher than central. The 95 % limits of agreement were -8.6 to 9.0 mmol/mol (local minus central). Eighty percent of local measurements were within ±10 % of corresponding central measurements. Some trial centres were more varied in the differences observed between local and central measurements: IQRs ranging from 3 to 9 mmol/mol; none indicated systematic bias.
Variation in agreement between HbA1c measurements was greater than had been expected although no overall bias was detected and standard deviations were similar. Discrepancies were present across all participating centres. These findings have implications for the comparison of standards of clinical care between centres, the design of future multicentre RCTs and existing quality assurance processes for HbA1c measurements. We recommend that centralised HbA1c measurement is preferable in the multicentre clinical trial setting.
Eudract No. 2010-023792-25 , registered on 4 November 2010.
糖化血红蛋白(HbA1c)是糖尿病临床试验中的一项重要疗效指标。对于多中心设计,HbA1c可在参与中心本地测量,或通过将血样送至中央实验室进行测量。本研究利用一项针对新诊断的I型糖尿病儿童的多中心随机对照试验(RCT)收集的1年随访数据,分析本地测量与中央测量之间的一致性。
在基线时以及随后的3、6、9和12个月,常规在本地和中央对HbA1c进行测量,并以mmol/mol报告数据。采用Bland-Altman分析方法,通过计算偏差和95%一致性界限来评估一致性。将HbA1c测量之间临床上可接受的误差范围的预定基准主观设定为±10%。计算被分类为临床上可接受的测量对的百分比。使用描述性统计来检查各中心内部的一致性。未考虑治疗组。
比较了代表25个5儿童和15个试验中心在四个随访时间点的590对测量值。无显著偏差:本地测量值平均比中央测量值高0.16 mmol/mol(标准差=4.5,95%置信区间-0.2至0.5)。95%一致性界限为-8.6至9.0 mmol/mol(本地测量值减去中央测量值)。80%的本地测量值在相应中央测量值的±10%范围内。一些试验中心在本地测量与中央测量之间观察到的差异更大:四分位数间距范围为3至9 mmol/mol;均未显示出系统偏差。
尽管未检测到总体偏差且标准差相似,但HbA1c测量之间的一致性差异大于预期。所有参与中心均存在差异。这些发现对各中心之间临床护理标准的比较、未来多中心RCT的设计以及HbA1c测量的现有质量保证流程具有影响。我们建议在多中心临床试验环境中,中央化的HbA1c测量更为可取。
Eudract编号2010-023792-25,于2010年11月4日注册。